具有长类和字段名的Java默认序列化

Question

my question is, how does java serialisation work? I'm serializing 100_000 objects of a class with a long class name and long field names. And it takes much more time then with another class with short class/field names. So does java serialization write the class and the field names with every object? Or maybe just the field names with every object?

Answer 1

Java writes the class descriptor once for each class in the serialization stream. The class descriptor includes the class name, the field names, and the class names (or references to class names) of the field types.

The class descriptor for an object is embedded in the object for the first object of that class in a stream. Subsequent objects of the same class only contain a reference to the class descriptor.

Field names only appear in the class descriptor. The object's field values are represented as an array of values; no names.

It is likely (though not obvious from the spec) that class and field names in class descriptors are de-duped.

For more details, refer to the Object Serialization Stream Protocol specification.

---

This means that long class and field names will make a serialization longer, but there should not be significant repetition of the names.

It takes much more time then with another class with short class/field names.

That is surprising if (as you seem to be saying) you are putting 100 objects of the same class into a stream.